Vascular bifurcation influences the protein corona composition on nanoparticles and impacts their cellular uptake

The protein corona (PC) that forms on nanoparticles (NPs) after in vivo injection influences their biodistribution, pharmacokinetics, and cell interaction. Although injected NPs traverse vascular networks, the impact of vascular features on the protein corona composition is mainly unexplored. Using an in vitro flow model that introduces bifurcations, a common feature of blood vessels, we show that vessels are not passive bystanders in the formation of the PC but that their features play active roles in defining the PC on NPs. The addition of bifurcation significantly increased the amount of proteins associated with NP. The bifurcation's introduction also changed the PC's composition on the NPs and affected the NP interactions with cells. Correlation analysis and modeling showed that these changes in the PC are mediated by both the branching and diameter reduction associated with vessel bifurcation and the resulting change in flow rate. The results indicate that blood vessel structures play an active part in the information of the PC, and their role should be studied critically for a better understanding of the PC and its biological implications.


Introduction
Nanoparticles (NPs) continue to see increasing use in medicine for drug delivery, medical imaging, and diagnosis. 1,2 NPs have tuneable properties such as size, shape, and surface chemistry that can be controlled for different medical applications. 3 Their large surface-to-volume ratio allows conjugation of ligands, thereby enhancing their delivery to target tissues for therapy and imaging. 4,5 Aer in vivo administration into biological uids, proteins rapidly adsorb on NPs, forming a protein corona (PC) due to their high surface free energy. 6,7 The PC imparts a new biological identity to the NPs that governs their biological fate. The PC composition inuences NP uptake by immune cells with implications for their pharmacokinetics and biodistribution. 8 It also affects cellular responses, including inammation, oxidative stress, and apoptosis. [9][10][11] The biological effect of the PC on NP fate is determined by the type and amount of the bound proteins, predicted by the NP physicochemical characteristics and the physiological conditions under which the PC is formed. 12,13 Physicochemical properties of NPs such as size, shape, surface charge, and surface functionalization directly mediate the formation of the PC and inuence the composition of the protein corona. 8,12,14 However, even with extensive studies, there remains discordance between the in vitro identied protein corona and those identied on NPs collected from in vivo studies for particles with the same physicochemical characteristics. 15 Recently, the effect of ow on PC composition and its ultimate impact on NP fate has been recognized, 16,17 with the ow velocity directly affecting the composition of the PC. 13 These results also demonstrate that shear stress generated under dynamic ow conditions affects the complexity and abundance of protein corona on NPs. 18 Studies comparing PC formed on NPs under dynamic ow and static conditions in vitro show variation in the amount and composition of the PC that affects NPs' cellular uptake and toxicity. 17 Characterizing the PC formed on intravenously injected NPs in vivo and comparing it to PC formed in vitro under ow conditions has further highlighted how the complex physiological environment inuences the composition of the PC and the resulting biological effect, independent of their physicochemical properties. 19 Understanding how physiological attributes of the vasculature other than ow velocity inuence the composition of the PC is necessary to reconcile the seeming disparity between PCs formed in vivo and in vitro. Resolution of this problem will enable better prediction of the PC composition on NPs and their potential biological effects.
Bifurcation is a common feature of blood vessels, with larger vessels progressively branching into smaller daughter vessels. 20 Such changes oen lead to variations in blood ow. Poiseuille's law states that the volume ow rate directly depends on the vessel's geometry, such as radius of curvature and length, and is inversely proportional to viscosity. 21,22 With the viscosity of the liquid remaining constant throughout, the geometry of the blood vessel can have a predominant effect on ow prole. 23 Therefore, constant branching and associated reduction in diameter could alter the ow prole of plasma. Turbulence is a crucial feature observed in plasma ow that arises due to changes in vessel physiology. 24 Such changes in plasma ow caused by vascular dynamics could potentially affect the distribution of proteins in the blood and their interactions with NPs. 25 A classic example of how vessel structure affects the ow and its implications is the increased probability of plaque depositions and leukocyte adhesion at vessel bifurcations. 26,27 Unfortunately, most studies only evaluate ow rate and do not consider vessel architecture.
This study analyses NP protein interactions in the presence of a single bifurcation with a decrease in the diameter of the daughter vessels to determine the impact on the composition of the PC that forms on polystyrene NPs. The results show that the changes in vessel structure indeed inuence the amount and composition of the PC on NPs. Changes in the NP uptake were also observed due to differences in the PC composition. These results have implications for the ability to translate in vitro PC studies in vivo and suggest the need to include the complexity of the vasculature in the design of in vitro studies.

Results and discussion
Modeling the vasculature in vitro to study effort on protein corona formation Blood vessels branch into smaller daughter vessels that change blood ow velocity. Considering that the ow rate of plasma affects the formation of PC on NPs, we hypothesized that changes in hemodynamics caused by bifurcations and the associated reduction in blood vessel diameter impact the composition of the PC formed on an NP. In vitro models made from polyurethane tubing connected to a peristaltic pump to form a closed-loop circuit or incorporating bifurcations (Fig. 1A) were created to test the hypothesis. For the bifurcation model, a central tubing with 3/32 inches internal diameter (ID) branched into two daughter tubes of 1/16 inch ID at an angle of 70 between them (Fig. 1A). Fluid dynamics were modeled for these vessel architectures with ANSYS Fluent soware to evaluate how hemodynamics are affected. The modeling of uid ow in the loop system had a maximum velocity of 2.9 m s À1 at the center that deceased signicantly closer to the vessel walls. This is expected as ow in cylindrical tubes has maximum velocity in the center, while the velocity decreases closer to wall due to boundary condition. 28 While there is no uid mixing in the simple loop, the addition of the bifurcations increased mixing, causing turbulence at the branch junction. In the models with bifurcation, uid velocities peaked at 3.1 m s À1 at the branch point, 6% faster than the highest ow rate predicted in the loop conguration. Immediately aer the branch point, the uid velocity drastically reduced within the arms of the branches (1.6 m s À1 ) before increasing to a maximum of 2.3 m s À1 at the center of the daughter vessels. These in silico modeling results show the effects that the dramatic changes in the vessel architecture could have on uid ow and corresponds to ow rate variations modeled for in vivo bifurcations. 29 Bifurcation affects the amount of proteins on NPs To investigate the potential effect of bifurcation on the protein corona, 50 nm amine-modied polystyrene NPs were purchased and used without any modications. Polystyrene NPs (PS-NP) are used as a model NP in these studies because their uniformity in size (polydispersity index ¼ 0.06) would reduce possible confounding results due to size distributions.
The hydrodynamic size of the PS-NPs was measured as 54 AE 1 nm, while the core size, derived from transmission electron microscopy (TEM), was 50 AE 10 nm ( Fig. 2 and Table 1). The PS-NPs were injected into circulating plasma in the loop and bifurcation systems at a ow rate of 25 mL min À1 for 10 minutes. The particles were washed thrice with 1x PBS to remove the loosely bound proteins (so PC), leaving behind proteins tightly bound to the surface called the hard PC. The resultant NP-hard PC complex was used for all the further analyses. The hydrodynamic diameter of the PS-NPs incubated in the loop system was 66 AE 27 nm, demonstrating an increase in the hydration shell due to the adsorbed hard PC. The TEM size remained unchanged ( Fig. 2 and Table 1). Aer incubation of the PS-NPs in the bifurcation model, the hydrodynamic diameter increased to 56 AE 18 nm, while the TEM measured size was 51 AE 8 nm ( Fig. 2 and Table 1). The zeta (z) potential of the untreated amine-modied PS-NPs was +42 AE 5 mV and changed to À22 AE 4 mV and À25 AE 2.4 mV aer incubation in the plasma in the loop and bifurcation systems, respectively ( Table 1).
The plasma proteins bind to the surface of NPs upon immediate contact with its surface, reducing the exposure of amines on the NP surface that are responsible for the cationic charge. The z potential reversal of the NPs upon exposure to plasma indicates the preferential binding of proteins with low isoelectric points (pI) to the cationic NP surface. Albumin, the most abundant protein in blood plasma and generally present in the PC has a pI of 4.9. 30,31 Similarly, complement proteins and apolipoproteins, which generally form part of the PC, have low pIs. 31 Although the TEM micrograph of the untreated PS-NP showed individual unaggregated particles, PS-NPs from the loop and bifurcation treatment showed the tendency to aggregate. However, the pattern for aggregation was different for both treatments. The PS-NP from the loop treatment aggregated in series to form chain-like structures (Fig. 2), while the PS-NPs from the bifurcation treatment aggregated into large clusters. The amount of protein on NPs inuence how they aggregate. 32,33 At optimal concentrations, proteins improve the distribution of NPs and prevent aggregation. However, higher concentrations of proteins on NPs tend to increase particle aggregation. 33 Proteins that bind to NP surfaces undergo conformation changes and could induce protein aggregation, enhanced by the increased turbulence in the bifurcation model. This can ultimately lead to coalescence of the protein-coated NPs into aggregates. These results suggest that different amounts of proteins with different compositions adsorb to the NPs aer treatment under the two conditions. This is anticipated as the ow rate of plasma has been shown to affect the amount of proteins on NPs. 17 We conrmed that the change in ow velocity caused by the introduction of the bifurcation impacted the amount of proteins in the hard corona by micro-BCA assay. The addition of the bifurcation increased the amount of protein on the PS-NPs by 75% compared to the loop system ( Fig. 3A; p ¼ 0.0324). These results are supported by other studies that show that NPs isolated aer in vivo injection have more adsorbed proteins than in vitro incubated NPs, supporting a role for the vascular architecture. 15,34 We next determined whether the difference in the amount of protein in the PC also translated into variations in its composition using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Four independent experiments were performed for each condition, and proteins with an abundance greater than 0.2% were analyzed. The protein had to be present in at least three experiments for it to be considered for further analysis.
Of the most abundant proteins identied by LC-MS/MS in the hard corona, 15 unique proteins were present only in the corona for NPs injected in the loop system. In comparison, 14 unique proteins were present in only the PC from the NPs injected into the bifurcation system. There were 32 proteins present in the PC from NPs treated in both the loop and bifurcation systems.
Although 32 proteins were shared between the two models, their abundances varied and were impacted by the ow models (Fig. 3B). Fig. 3C represents the 10 proteins that showed signicant differences in abundance between the two  conditions. While overall, bifurcation increased the amount of proteins in the PC, the shared proteins were most abundant in the PC from the loop condition (Fig. 3C).

Bifurcation changes the composition of the protein corona on PS-NPs
The proteins were categorized depending on their contribution to acute phase reactions, coagulation, complement response, immunoglobulins, and lipoproteins. These protein groups were selected because of their contribution to NP biodistribution, pharmacokinetics, clearance, and cellular interactions. 19,34 For example, lipoproteins, acute phase reactants, immunoglobulins, and the exclusion of coagulation and complement proteins in the PC of injected NPs increase their blood residence time. 34 48.3% and 36.7% of the LC-MS identied proteins fell in the categorized groups for loop and bifurcation treatments, respectively (Fig. 4A). Of these proteins, lipoproteins were most abundant, representing 18% and 13% in abundance in the PC on NPs from the loop and bifurcation ow systems, respectively. Apolipoprotein (Apo) B-100 was the most abundant lipoprotein accounting for more than 50% of the lipoprotein content in PC of NPs from both loop and bifurcation models. While present in both PCs, clusterin, Apo A-I and Apo C-III demonstrated statistically signicant differences in abundance ( Fig. 4B and  3C). Proteins associated with coagulation were most prominent aer lipoproteins and showed higher abundance in the PC of NPs from the loop system. The abundance of brinogen components (alpha, beta, and gamma chains) was twice in the PC of NPs from the loop system compared to the PC from the bifurcation system. The PC from the loop also lacked alpha-2antiplasmin and kininogen-1, which were present in PCs from bifurcation, albeit at low abundance. The composition of complement system-related proteins in the PC of NPs from the bifurcation system was more diverse, with 10 different proteins compared to just 5 in the PC from the loop system. Similarly, there were differences in the identity of proteins related to acute phase reactions and immunoglobulins (Fig. 4B-F).
Considering that the introduction of bifurcations affected the PC composition on the NPs, the potential impact on biological pathways was investigated. For this, the annotated LC-MS/MS data were analyzed using Reactome Pathway Database, a free, open-source bioinformatics tool that maps pathways by analyzing the over-representation of proteins associated with specic pathways in a given data set. [35][36][37] The analyses show enrichment of proteins associated with different pathways in the PC on the NPs (Fig. 5). Although there was equal enrichment   of proteins associated with platelet activation, aggregation, and degranulation in the PC of NPs from both ow systems, proteins associated with brin clot formation were more enriched in the PC from the loop. These together resulted in pathways associated with hemostasis being enriched in the PC on NPs from the loop system. Pathways associated with complement activation and progression were more abundant in the PC from the NPs that owed through the bifurcation model. However, pathways contributing to insulin-like growth factor protein transport were enriched in the loop model.

The PC formed on NP due to bifurcation disrupt cellular uptake
The PC mediates NP cellular interaction and uptake endocytosis. 38 The PC-coated nanoparticles (PC-NPs) were incubated with 4T1 breast cancer cells to assess cellular uptake and toxicity. Both MTS metabolic assay and lactate dehydrogenase (LDH) viability assay show that bare PS-NPs are toxic to the 4T1 cells aer incubation for 24 hours in serum-free conditions ( Fig. 6A and B). The presence of the protein corona reduced NPinduced toxicity, which was independent of how the PC was formed ( Fig. 6A and B). The toxicity of many NPs is mediated by their surface chemistry. [39][40][41] Cationic NPs such as these PS-NPs induce toxicity by disrupting the cell membrane, with their toxicity rising with increasing z-potential. 40,42 The PC on the NPs masks this cationic surface and prevents direct interaction with the cell membrane. The PC-coated NPs (PC-NP) were incubated with 4T1 cells for 4 hours. The uptake studies were performed in serum-free media to ensure that any observed effect is from the proteins associated with the NPs. The NPs with PC from the loop system showed uptake by the 4T1 cells, which were distributed throughout the cytoplasm, and did not show a preference for perinuclear localization (Fig. 6C). However, the NPs with PC generated in the bifurcation ow system showed minimal uptake by the 4T1 cells. The NPs mainly remained outside the cell, associated with the cell membrane as clusters (Fig. 6C). Our initial TEM analysis of the PC-NPs showed that NPs incubated in the bifurcation system tended to form clusters (Fig. 2B), likely due to the greater amount of proteins in the PC. The increased hydrodynamic shear forces experienced by NPs and proteins in the bifurcation model could also cause protein unfolding and denaturation that increase aggregation. The cellular uptake of NPs is a function of their size, with NPs larger than 100 nm showing lower uptake. 42,43 Although the individual NPs are 50 nm and readily endocytosed, their aggregated size of greater than 500 nm decreases their endocytosis by the cells. While this aggregation may hinder their uptake by 4T1s, which are of epithelial origin, they may increase uptake by macrophages that remove debris. 44,45 These results suggest the probability that the NPs would aggregate and accumulate in the liver and spleen, lowering their blood residence time in vivo.

The protein corona is impacted by both branching and size reduction
The results obtained showed that bifurcation indeed affects the PC formed on NPs. However, they raise the question of what component of the bifurcation, the branching, reduction in diameter of the daughter vessels, or combination of both is responsible for the observed changes. The PC formed on the PS-NPs aer ow in two other ow systems; a bifurcation model without the decrease in the diameter of daughter vessels and a loop system that incorporates a decrease in size were characterized to determine this (Fig. 7A). Dynamic ow simulations show that decreasing the diameter of the vessel increased uid velocity signicantly, reaching speeds of 6.3 m s À1 at the center of the vessel. In the branching model without size reduction, in the portion of daughter vessels, there is an increase in velocity at the branch point that is followed by a decrease within the branch. However, the decrease in velocity is not as dramatic as observed in the standard bifurcation model with a decrease in the diameter of daughter vessels. The composition and amounts of proteins, as abundance percentage, in the PC on NPs from these two models were compared to those from the standard bifurcation model (Fig. 7C). Comparison of PC of NPs from bifurcation with size reduction to those from bifurcation without size reduction in daughter vessels yielded an R 2 ¼ 0.76. Comparison of the PC from the bifurcation with size reduction to those from the loop with size reduction produced an R 2 ¼ 0.85. These results demonstrate that the PC composition on NPs results from a combination of the branching and reduction in the size of the daughter vessels.

Conclusion
The study presented here demonstrates the impact of bifurcation, a common feature of blood vessels on the PC formed by NPs. Bifurcations increase the amount of and change the composition of proteins adsorbed to NPs and ultimately determine their interaction with cells. This is the rst study to explore how different aspects of blood vasculature could contribute to the formation of the PC on injected NPs. Considering the current discordance between PC formed on NPs in vitro and in vivo, this study suggests the varied results may be partly due to the vessel structure and its impact on uid dynamics. Therefore, there is a need to investigate better how vascular characteristics impact the PC to understand their formation in vivo better and create better prediction models.

Materials and methods
Materials 50 nm blue uorescent amine-modied polystyrene NPs were purchased from Millipore Sigma (Burlington, MA) and were used without further modication. Human plasma with K2 EDTA anticoagulant was purchased from Innovative Research (Novi, MI) and centrifuged to remove aggregated protein mass. Polyurethane tubing and ttings of desired sizing and shape were purchased from United States Plastic Corps (Lima, OH).

Protein corona formation on nanoparticles
Flow conditions for plasma proteins were established using a peristaltic pump. Polyurethane tubing of desired geometry was constructed using appropriate ttings and connected to the pump. The bifurcation model was created using polypropylene tting with a 70 angle between the two arms. To mimic an artery with branching followed by a reduction in diameter, polyurethane tubing with an inner diameter of 3/32 inner diameter (ID) was attached to two smaller 4 inches diameter tubing of 1/16 inches ID via the "Y" tting. The smaller tubing ended in another "Y" tting connected to a larger diameter tubing to create a closed circuit vessel model. The simple loop model was constructed using silicone tubing of the equivalent length of the bifurcation circuit. The volume of plasma in both systems remained equal at 2 mL.
The models were lled with 100% human plasma (2 mL) and set to a 25 mL min À1 ow rate using a peristaltic pump (Fish-erbrand™ Variable-Flow Peristaltic Pumps). Using an insulin syringe, polystyrene NP(1 mg mL À1 ) was injected in the plasma ow direction and allowed to circulate for 10 minutes at 37 C. Aer 10 minutes, the NPs were recovered from the plasma by centrifugation (Thermo Scientic™ Sorvall™ Legend™ Micro 21 Microcentrifuge) at 21000 g for 30 minutes. The recovered NPs were washed in phosphate-buffered saline (PBS) three times to remove the so corona, with centrifugation recovery aer each wash.

Particle characterization
Transmission electron micrography (TEM) (JEOL 1200EX) conrmed the size distribution and morphology of plasmatreated and untreated nanoparticles. Nanoparticles were dropcast onto copper grids, stained with uranyl acetate, and airdried at room temperature before imaging.
Hydrodynamic diameter, polydispersity index, and zeta potential measurements were performed on a Malvern Zetasizer-Nano instrument on particles resuspended in ultrapure water at room temperature.
The amount of protein on NP pellets aer running through the ow models was obtained aer washing three times with 1x PBS by micro-BCA assay (ThermoFisher, Waltham, MA) following the manufacturer's instructions.

Mass-spectrometry
Sample preparation. Protein digestion was carried out by S-Trap sample processing technology (S-TRAP™) according to the protocol. Briey, reductant (1 mL) was added to protein (5 mg) and was dissolved in 1X lysis buffer. Aer incubation for 15 minutes at 55 C, an alkylator (1 mL) was added, and the solution was incubated for 10 minutes. The pH of the solution was adjusted to 1 with acid and digested overnight at 37 C. The peptides were eluted using elution buffer, dried, and resuspended to the desired LC-MS/MS analysis volume.

LC-MS/MS
LC-MS/MS analysis used a Thermo Scientic UltiMate 3000 nanoUHPLC system coupled to an Orbitrap Fusion mass spectrometer. Dried samples were reconstituted in a mixture containing 98% water, 2% acetonitrile, and 0.1% formic acid (20 mL) and the injection volume was set to 1 mL. Samples were separated with an Acclaim PepMap C18 column (0.075 Â 150 mm, 2Å particle size) at a ow rate of 0.400 mL min À1 . The mobile phase consisted of water, acetonitrile, and formic acid (98/2/0.1 for buffer A, 2/98/0.1 for buffer B). The gradient was as follows: equilibration at 2% B for 5 minutes, ramping up to 45% B at 37 minutes, 90% B at 40 to 46 minutes, followed by reequilibration at 2% B at 47 to 60 minutes.
Eluted peptides were introduced into the mass spectrometer by positive ESI at 2450 V, and the ion transfer tube temperature was set to 275 C. Data was obtained in top speed mode with a cycle duration of 3 seconds. Full scans were acquired in the Orbitrap at a resolution of 120 000 at m/z 200. The mass range, RF lens amplitude, and maximum injection time were 400-1600 m/z, 60%, and 100 ms, respectively. MS/MS spectra were acquired in the ion trap set to a rapid scan rate. The precursor isolation window was set to 1.6 m/z. Fragmentation was achieved by HCD at a xed collision energy of 28%. Dynamic exclusion was set to 60 seconds with a tolerance of 10 ppm.
Data were processed by use of Thermo Scientic Proteome Discoverer 2.4. A custom theoretical fragment database was derived from the human reference proteome available from UniProt (UP000005640, 78120 protein sequences). Spectra were processed and matched against the database using the basic SequestHT processing and consensus workows included with the soware and tailored for the tribrid platform, with the precursor and fragment tolerances set to 5 ppm and 0.6 Da, respectively. Two peptides were required for the successful identication of a protein. The abundance percentage of each protein was calculated by dividing each peptide number by the sum of peptides in that particular group. The abundance percentage thus calculated was used for all the analyses.
Cell culture 4T1 breast cancer cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA). The cells were grown and maintained in RPMI supplemented with 10% fetal bovine serum (v/v), 1% penicillin-streptomycin in a 5% CO 2 incubator at 37 C.
The 4T1 cell lines were passaged at 70-80% conuency and were seeded in a 96 well-plate, with each well comprising 10 000 cells. The cells were allowed to attach to the well-plate prior to administration with the nanoparticle.

Viability assay
Cell viability was assessed 24 hours aer incubation with bare NPs or NPs with PC using the Lactate Dehydrogenase assay (CytoTox 96® Non-Radioactive Cytotoxicity Assay) and MTS assay (CellTiter 96® AQueous Non-Radioactive Cell Proliferation Assay) kit, following manufacturer's protocols.

Cell uptake of NPs
Aer incubation in plasma under the different ow conditions and washing in PBS to remove the so corona, the PC coated PS-NPs were uorescently labeled with Alexa Fluor 594 NHS ester (ThermoFisher Scientic, Waltham, MA). The NPs were sonicated for 5 minutes in a bath sonicator (VWR® Ultrasonic Cleaners) to break down the aggregates before addition to cells.
The protein corona coated NPs (25 mg mL À1 ) were incubated with 4T1 cells at a seeding density of 10 000 cells per well in a 96 well-plate, in serum-free media for 4 hours. The cells were washed three times with phosphate-buffered saline (PBS), stained with Hoechst 33258 dye, and xed in 4% paraformaldehyde before imaging.

Cell imaging
The 4T1 cells were imaged using a Lionheart LX automated microscope (Biotek, Winooski, VT) using the lter cubes corresponding to Texas red channel to image NPs and DAPI channel to image cell nuclei.